Download PDF file for New Tools for Labor Market Analysis:JOLTS

Job Openings and Labor Turnover
Job Openings and Labor Turnover
New tools for labor market
analysis: JOLTS
As a single, direct source for data
on job openings, hires, and separations,
the Job Openings and Labor Turnover Survey (JOLTS)
should be a useful indicator of the demand for labor
in the U.S. labor market and of other economic conditions
Kelly A. Clark
and
Rosemary Hyson
Kelly A. Clark is an
economist and
Rosemary Hyson is a
research economist in
the Office of
Employment and
Unemployment
Statistics, Bureau of
Labor Statistics.
32
A
nalysis of the U.S. labor market is a difficult and challenging effort. A number of
existing economic indicators, including
the unemployment rate, payroll employment, and
others, serve as useful measures of labor market
activity, general economic conditions, and labor
supply. However, to facilitate a more comprehensive analysis of the U.S. labor market and to show
how changes in labor supply and demand affect
the overall economy, the Bureau of Labor Statistics will introduce a new data series measuring
labor demand and turnover: the Job Openings
and Labor Turnover Survey (JOLTS).1
The availability of unfilled jobs—the number
of job openings or the openings rate—is an important measure of the tightness of labor markets. JOLTS calculates the number of job openings
from a nationwide sample of establishments and
computes a job openings, or vacancy, rate. This
new survey also collects data on separations by
type and hires, providing a single source for
these data that will enhance empirical analyses
of the economy and the labor market. This article
briefly describes the survey and then discusses
how JOLTS data will help enrich analysis of the
U.S. labor market and the economy as a whole.
The survey
collects and analyzes monthly data on many
aspects of the U.S. labor market. One BLS survey,
the Current Employment Statistics (CES) survey,
BLS
Monthly Labor Review
December 2001
collects data from businesses and produces employment estimates. Another survey, the Current
Population Survey (CPS), conducted by the Census Bureau, collects employment status data from
households for BLS to determine the unemployment rate, which measures excess labor supply.
The BLS Job Openings and Labor Turnover Survey, completes the labor market picture by collecting monthly data from businesses to measure
unmet labor demand and job turnover.
This new program involves the collection, processing, and dissemination of job openings and
labor turnover data from a sample of 16,000 business establishments. The universe frame for the
JOLTS sample consists of approximately 8 million
establishments compiled as part of the operations
of the BLS Covered Employment and Wages, or
ES-202, program. This frame includes all employers subject to State Unemployment Insurance (UI)
laws and Federal agencies subject to the Unemployment Compensation for Federal Employees
(UCFE) program. The sampling frame is stratified
by ownership (public or private), geographic region, major industry division, and size class. The
JOLTS sample is representative of private nonfarm
establishments as well as Federal, State, and local government entities in the 50 States and the
District of Columbia. The sample is rotated so
that most establishments participate in the survey for 18 consecutive months. Total employment
estimates from JOLTS are controlled to the current
month CES employment estimates, and this is used
to adjust the levels for all other JOLTS data elements.
The data elements collected monthly from each establishment include employment for the pay period that includes the
12th of the month; the number of job openings on the last
business day of the month; and hires, quits, layoffs and discharges, and other separations for the entire month. To encourage consistent and accurate reporting, respondents are
given detailed definitions for each data element. For example,
the definition of a job opening requires that a specific position exists, the job could start within 30 days, and that the
employer is actively recruiting from outside the establishment
to fill the position. Hires are all additions to the payroll during
the month, and a layoff should be counted if it lasts or is
expected to last more than 7 days.
BLS anticipates releasing monthly estimates of job openings,
hires, and separation rates and levels beginning in early 2002.
The JOLTS data series will be considered a developmental series
for the first 2 years of publication. Estimates will be released for
the Nation as a whole and for four geographic regions. The
national estimates for the private sector will be divided into nine
industry divisions based on the Standard Industrial Classification (SIC) system. Additional estimates will be published for the
Federal Government and for State and local governments combined. JOLTS estimates will be converted to the North American
Industry Classification System (NAICS) in 2003.2
Anticipated uses of the data
The JOLTS data series on job openings, hires, and separations
will assist policymakers and researchers in addressing some
fundamental issues concerning labor demand and movements
in the labor market. The JOLTS data will provide a basis for
improved understanding of the factors driving fluctuations in
unemployment and the overall economy, determining appropriate approaches for reducing unemployment, and studying
how workers flow in and out of establishments, are matched
with jobs, and are distributed across sectors.
Of particular interest to researchers will be the study of the
relationship of vacancies and unemployment. The JOLTS job
openings rate is expected to have a negative relationship to
the unemployment rate. For example, as the economy expands,
the unemployment rate generally falls, reflecting a decreased
pool of excess workers. Simultaneously, the job openings rate
is expected to increase as businesses seek workers to fill new
and existing jobs. Alternatively, in a weak economy, the unemployment rate typically increases as businesses shed workers
in response to weak demand, and the job openings rate is
expected to fall as employers cut back their hiring plans.
The empirical relationship linking vacancies with unemployment and the overall economy is called the Beveridge
curve. A stylized version of the Beveridge curve is shown in
exhibit 1. First described by William Beveridge in the 1940s,
the curve reflects the negative relationship between vacancies and unemployment.3 Changes in the business cycle generate movement along the curve, while reallocation of labor
and capital between sectors moves the curve further away
from or closer to the origin. These forces are not independent
of one another, and changes in labor force composition and
job search behavior also affect where the Beveridge curve is
relative to the origin. Economic researchers can use the JOLTS
data to test and refine models that show how such factors
affect labor market dynamics and the distribution of workers
across sectors.
Another topic of importance to economists is the relationship between changes in wages and changes in employment.
Primarily, economists have only been able to examine this in
the context of the Phillips curve, a relationship linking deviations from the natural rate of unemployment—or excess labor supply—to wage changes. The JOLTS series on job openings will finally provide a consistent series on excess labor
demand that will extend economists’ ability to study the relationship between wage and employment changes.
To study the Beveridge curve and its implications, researchers have had to use a variety of proxies for the vacancy rate.
Much work in this area has been done using a vacancy series
constructed from employment, as measured by BLS, and the Conference Board’s Help-Wanted Advertising Index. This proxy has
served as an imperfect measure, because the index does not
cover the entire country, job openings are not directly counted,
there are no precise definitions or adjustments for what constitutes actual job openings, and there are other factors independent of labor demand that can increase the volume of want ads.
JOLTS will provide a directly measured vacancy series that can
be used for this type of analysis.
The Bureau of Labor Statistics has conducted various surveys to measure job openings in the past, but these were
either short lived or were limited to the manufacturing industry. JOLTS was developed to consistently record monthly job
openings for all nonagricultural industries, including both
public and private businesses, and is representative of the
national economy.
Researchers have used the available proxies to study
movements in the unemployment-vacancy relationship over
the past few decades because a consistent set of vacancy
data has not been available. Generally, this work has documented upward movement along the Beveridge curve during
the 1960s, and a combination of movement along the curve
and outward shifts in the relationship for various periods
during the 1970s and 1980s.4 Recent work by Hoyt Bleakley
and Jeffrey C. Fuhrer documents an inward shift of the curve
in the early 1990s.5
Uses of JOLTS data in economic analysis. The analysis of
the co-movement of the unemployment and vacancy rates
Monthly Labor Review
December 2001
33
Job Openings and Labor Turnover
can provide more information about business cycles and the
type of unemployment that is prevalent in the economy. When
the structure of the economy is fixed, the position on the
Beveridge curve reflects whether an economy is expanding or
contracting, as shown in exhibit 1. In times of economic expansion, the unemployment-vacancy combination will be high
on the curve (low unemployment rate, high vacancy rate), and
in contractions it will be low (high unemployment rate, low
vacancy rate). The changes in unemployment generated by
this movement will mostly reflect cyclical fluctuations in labor
demand. The Beveridge curve will shift with changes in the
probability that jobseekers and job openings will be matched,
as shown in exhibit 2. The curve will shift away from the origin
if various factors make it harder for unemployed workers to fill
vacancies and shift back when matching efficiency improves.
These shifts reflect changes in structural unemployment or
frictional unemployment, or both, that result from changes in
the composition of the labor force, from reallocation of jobs
across sectors or geographic areas, or from other changes in
the way workers match with jobs. In considering the curve’s
shape and movement of the unemployment and vacancy rates
over time, economists can determine the differences between
unemployment driven by deficient demand and unemployment generated by reduced matching.
For example, suppose the unemployment rate begins to
rise substantially. If no corresponding decrease in the vacancy rate is observed, this points to an outward shift in the
Beveridge curve and a reduction in matching efficiency. The
matching inefficiency could reflect structural unemployment
if the individuals who are without a job do not have the right
skills to fill the vacancies. To reduce unemployment, training
could be provided to workers who lack needed skills. If aggregate shocks are driving the increase in unemployment—that
is, there is movement along the Beveridge curve—efforts to
train workers will produce little improvement. Rather, macroeconomic efforts to spur job creation may be more effective.
The models that distinguish between the amount of a
change in the unemployment rate attributable to cyclical fluctuations and the amount generated by factors that reduce
matching efficiency are well described by a number of authors. Katharine G. Abraham and Lawrence F. Katz and Olivier
Blanchard and Peter Diamond demonstrate how data on vacancies can differentiate between deficient demand and unemployment generated by reduced matching.6 An article by
Barbara Petrongolo and Christopher A. Pissarides discusses
issues regarding the matching function—a relationship that
describes how workers match with jobs and which relates
changes in hires to changes in vacancies and unemployment.7 A later section of this article outlines some of the basic
ideas behind these models.
JOLTS elements as indicators of the business cycle. As described earlier, the job openings rate, in conjunction with the
34
Monthly Labor Review
December 2001
Exhibit 1.
Movement along the Beveridge curve
Vacancy rate
Expansion
UL, VH
Contraction
UH , VL
Unemployment rate
unemployment rate, provides information about the state of
the economy. In addition, hires, quits, and layoffs and discharges data can be useful in analyzing business cycles. Direct measures of all these data series have not been available
in the past. Many hypotheses about how these data elements
trend against one another and how they relate to the movements of the business cycle over time can be tested once the
JOLTS data are available. Economists will be able to examine
how the data elements move with business cycles directly. In
addition, economists can use the JOLTS series to enhance analyses of how gross employment flows and job creation and job
destruction relate to the business cycle.
Movement in the job openings (or vacancy) rate leads
economic activity at business cycle peaks and lags at troughs,
so that the number of vacancies tends to decrease before the
economy begins a downturn and increases after the economy
begins a recovery. Because cutting job openings is less costly,
employers tend to reduce the number of job postings for new
hires or replacement workers before decreasing the firm’s current employment level when sensing an economic downturn.
When conditions begin to improve, firms will tend to increase
hours of work and recall workers from layoff before searching
for new employees.
Hires, as well as establishment growth rates, tend to be
procyclical, moving in the same direction as general economic
activity. Under good economic conditions, firms replace workers who separate, whereas during downturns, employers may
delay hiring until the economic situation improves.
Intuitively, economists expect quits to be procyclical if
workers feel safe leaving their jobs when economic conditions are favorable. It also is thought that layoffs and discharges are countercyclical, moving in the opposite direction
as economic activity, because employers tend to shed workers when business conditions are unfavorable. Patricia M.
Anderson and Bruce Meyer found total separations, which
includes quits and layoffs and discharges, to be procyclical.8
If quits are procyclical, they will decrease during downturns and
recessions. Therefore, if total separations are also procyclical, in
order for both quits and separations to fall during a downturn,
the decrease in quits must more than offset the increase in layoffs and discharges during those periods. However, this aspect
may vary by industry, because sensitivities to business cycles
can vary. Hoyt Bleakley, Ann E. Ferris, and Jeffrey C. Fuhrer
provide evidence that flows between employment and unemployment in manufacturing vary much more with the business
cycle than in the nonmanufacturing sector.9
Economists have examined how business cycles affect
workers moving between jobs, becoming unemployed, and
dropping out of the labor force. JOLTS does not measure such
flows, but its turnover data should enhance estimates of these
flows using other data sources. For example, if quits and hires
are procyclical, then employer-to-employer flows should be
procyclical. Recent work by Bruce C. Fallick and Charles A.
Fleischman, however, found no evidence that such employerto-employer flows are procyclical using State-level variation
in economic conditions during 1994–2000.10 The evidence on
the procyclicality of hires and quits tends to come from data
collected during the 1970s and 1980s; the new JOLTS series will
be able to help economists understand whether these relationships still hold or if other components need to be incorporated into models of how job turnover and labor market flows
relate to the business cycle.
Many of the conclusions that economists have reached
about the relationship between the business cycle and job
openings, hires, and separations have been based on data
series that are not direct measures of these data elements. The
JOLTS program’s data series will provide a better source for
economists to examine these relationships over time.
Uses of JOLTS in economic research. In a simple model in
which an economy has fixed structural characteristics, fluctuations in aggregate demand generate movements along the
Beveridge curve as the economy expands and contracts. During contractions, there are few vacancies and high unemployment rates; as the economy expands and demand for labor
increases, unemployed workers will find jobs and growing
firms will be creating new jobs, generating a higher vacancy
rate and lower unemployment rate.
The reallocation of economic activity across industries
and geographic areas also generates distinct movements in
unemployment and vacancies that will be reflected in shifts of
the Beveridge curve. Expanding sectors have greater employment growth and shrinking sectors experience reductions in
employment. If workers who lose their jobs in the shrinking
sectors are perfectly mobile and have the skills to work in the
expanding sectors, then reallocation across sectors would not
affect the position of the Beveridge curve. Such seamless
transitions are rarely the case; impediments such as the lack
of skills or geographic immobility affect jobseekers’ ability to
match with job openings. As a result, concurrent increases in
both vacancies and unemployment can be observed and shift
the Beveridge curve further away from the origin.
Demographic changes in the labor force also will affect the
degree to which workers match with jobs. Workers have much
higher turnover early in their career, so a change in the age
distribution of workers towards younger workers can generate shorter job durations. Shorter job durations are usually
associated with greater turnover and more openings at any
one time. Concurrent increases in both unemployment and
vacancies could result, shifting the Beveridge curve further
away from the origin. Researchers have found some evidence
that the Beveridge curve shifted out when the baby-boom
Exhibit 2. A shift in the Beveridge curve
Vacancy rate
Reduced
matching efficiency
Improved
matching
efficiency
Unemployment rate
Monthly Labor Review
December 2001
35
Job Openings and Labor Turnover
generation entered the labor market and shifted the distribution of the labor force towards younger workers. An anticipated question that economists may be able to examine with
the JOLTS data is: What effect will the retirement of the babyboom generation have on the unemployment-vacancy locus?
In addition, the entry of their children into the labor force is
also likely to be currently affecting the age distribution of the
work force and possibly shifting the Beveridge curve as well.
Others have noted how changes in family structure, work
disincentives, and changes in recruiting practices on the part
of employers can influence the number of vacancies relative
to unemployment.11 In addition, the increase in the flow of
information on job openings, made possible by electronic communication and the Internet, may reduce both structural and
frictional unemployment, reduce the duration of job vacancies, and thus shift the Beveridge curve inwards. Such questions cannot be directly examined using the JOLTS data, but its
job openings, hires, and separations data could be combined
with other microdata to investigate such questions.
Aggregate demand shocks, changes in the distribution of
jobs across industries and regions, shifts in the demographic
characteristics of the work force, and other changes in the
way labor markets operate all can affect the combination of job
openings and unemployment that are observed at any one
time. Economists are interested in the extent to which each of
these factors matter, as well as how long their impact persists.
Several researchers used the job vacancy series from the 1970s
to estimate the relative importance of each of these on the
movement of unemployment and job openings.
In looking at the evidence using earlier vacancy data series, researchers find that aggregate demand shocks generate
larger short-run movements in the vacancy-unemployment relationship than reallocation or changes in labor supply.12 The
impact of aggregate demand does not persist in the long run,
while the effects of changes in the labor force composition
and increases in the intensity of reallocation do.
A number of authors have found that the Beveridge curve
shifted out in the 1970s and 1980s. Changes in the composition of the labor force appear to have played at least a small
part in this shift.13 Reallocation of jobs across industries has
played a role, but these are not entirely of a consistent timing
and magnitude to explain the remainder of the shifts between
the 1970s and 1990s. One study provides evidence that increased geographic dispersion of job creation and destruction are more consistent with the shifts; all the researchers cite
additional factors that also were likely to have affected matching and shifted the Beveridge curve.14
JOLTS will be the first data series in which macroeconomic
analyses can be performed without constructing at least one,
if not more, of the main components of these models from
proxies based on manufacturing data. The studies from the
1980s had to rely on relationships between employment
36
Monthly Labor Review
December 2001
changes, job openings, quits and layoffs constructed from
turnover, and vacancy data collected in 1981 and earlier, all of
which were limited to the manufacturing industry. In addition,
these relationships were used in conjunction with household
data to approximate the number of persons moving from one
employer to another to generate measures of job separations
and hires that did not come from nonemployment. These constructed series on vacancies, quits, and hires seemed to fit
labor market data well through the 1980s.15 However, several
changes in the labor market over the past 30 years, including
the shift in the concentration of jobs from manufacturing to
services and changes in the way workers match with job openings, limit the extent to which one can hope to “update” these
earlier series. Moreover, it is always preferable to have direct
measures of a model’s components rather than having to construct some components. The hires series from JOLTS will provide data on the number of workers who match with jobs
during each month. The separations series and breakouts of
quits, layoffs and discharges, and other separations will provide the necessary turnover data and job openings will provide a direct measure of vacancies for these models.
The JOLTS data also will enhance economists’ understanding of matching models—the process by which employed and
unemployed jobseekers match with available jobs. The hires
series provides a monthly estimate of the number of such
matches, and the job openings series provides a measure of
the number of unfilled jobs. In analyses to date, economists
have had to rely on less than ideal measures of these two
critical elements. Often, exits from nonemployment were used
to proxy for job matches even though recent evidence indicates that a substantial number of workers go from one job to
the next without a spell of nonemployment. Also, these models have had to rely on proxies for a vacancy series or gross
employment changes to control for labor demand. The proxies for vacancies and job matches have been sufficient for
certain analyses, but their being imperfect measures of the
true values can confound testing of other hypotheses.
An example of how measurement error interferes with analyzing matching models can be seen in economists’ examination of the effect that increasing the number of jobseekers has
on the rate at which vacancies are filled. Some models imply
that at the aggregate level, the matching function should exhibit constant returns to scale—that is, the overall match rate
will equal the sum of the probabilities that any given jobseeker
and firm will match. In regression models that estimate the
elasticity of job matches or new hires with respect to vacancies and unemployment, constant returns to scale implies that
the elasticity of matches with respect to vacancies and the
elasticity of matches with respect to unemployment should
sum to one. Other models suggest that there might be increasing returns to scale in the matching function—that is, the
total number of matches will increase by more than the amount
that the number of jobseekers and job openings has increased.
Increasing returns to scale would imply that the estimated sum
of the unemployment and vacancy elasticities is greater than
one. Measurement error in either unemployment or vacancies
can result in the estimated elasticities being biased towards zero.
Therefore, imperfect measures of job vacancies might suggest
constant returns to scale simply because the estimated effect of
job vacancies is biased downward.
Matching models have been used to examine the extent to
which employers prefer employed jobseekers to unemployed
jobseekers. Such ranking of types of workers has implications for jobless durations and unemployment durations.
Recent evidence indicates that a substantial proportion of
new employees come from other jobs rather than unemployment.16 Although JOLTS does not collect data on whether hires
come from another job, unemployment, or out of the labor
force, its information on matches (hires) and job openings
will provide better information than existing proxies created
from other surveys.
With data from a single consistent source, there is less
concern that measurement error in the key components will
interfere with estimating the true relationship between vacancies, unemployment, and other economic indicators.
These series will improve the ability to test how well current
models describe the evolution of labor demand and labor
market dynamics in the United States. Also, the time series
on hires and quits, in combination with data on gross employment flows, may enable researchers to back out other
turnover series, such as employment-to-employment transitions, and further enhance such research.
These topics are part of economists’ research agendas
and extend beyond JOLTS use as an indicator of labor demand
in the United States. The full scope of the data’s usefulness
cannot be anticipated; however, as the data series accumulates and models of labor dynamics are further developed,
they may substantially deepen economists’ understanding
of labor markets and the economy.
AS A SINGLE SOURCE for directly measured data on job openings, hires, and separations, JOLTS statistics can be used as
indicators of general economic conditions, and are important
tools for considering the implications of economic policies
on unemployment and the labor market. These data series, in
conjunction with other micro- and macroeconomic data, are
also likely to enhance researchers’ understanding of labor
market dynamics and their relation to the economy as a whole.
Notes
1
For a brief introduction to the Job Openings and Labor Turnover
Survey see, “New survey measures demand for labor,” Monthly Labor
Review, September 2000, pp. 37–39.
8
Patricia M. Anderson and Bruce D. Meyer, “The Extent and
Consequences of Job Turnover,” Brookings Papers: Microeconomics,
1994, pp. 177–248.
2
See James A. Walker and John B. Murphy, “Implementing the
North American Industry Classification System at BLS,” also in this issue.
9
Hoyt Bleakley, Ann E. Ferris, and Jeffrey C. Fuhrer, “New Data
on Worker Flows During Business Cycles,” New England Economic
Review, July/August 1999, pp. 49–76.
3
See Hoyt Bleakley and Jeffrey C. Fuhrer, “Shifts in the Beveridge
Curve, Job Matching, and Labor Market Dynamics,” New England
Economic Review, September/October 1997, pp. 3–19. Also a derivation of the Beveridge curve can be found in Bent Hansen, “Excess
Demand, Unemployment, Vacancies and Wages,” Quarterly Journal of
Economics, 1970, vol. 84, no. 1, pp. 1–23.
4
See Katharine G. Abraham, “Structural/Frictional vs. Deficient
Demand Unemployment: Some New Evidence,” American Economic
Review, 1983 vol. 73, no. 4, pp. 708–24; and Katharine G. Abraham,
“Help-Wanted Advertising, Job Vacancies, and Unemployment,”
Brookings Papers on Economic Activity, no. 1, June 1987, pp. 207–48;
Olivier Blanchard and Peter Diamond, “The Beveridge Curve,”
Brookings Papers on Economic Activity, no. 1, 1989, pp. 1–76; and
Bleakley and Fuhrer, “Shifts in the Beveridge Curve” 1997, pp. 3–19.
5
Bleakley and Fuhrer, “Shifts in the Beveridge Curve,” 1997,
pp. 3–19.
6
Katharine G. Abraham and Lawrence F. Katz, “Cyclical Unemployment: Sectoral Shifts or Aggregate Disturbances?” Journal of Political Economy, 1986, vol. 94, no. 3, pp. 507–22; and Blanchard and
Diamond, “The Beveridge Curve,” 1989, pp. 1–76.
7
Barbara Petrongolo and Christopher A. Pissarides, “Looking into
the Black Box: A Survey of the Matching Function,” Journal of Economic Literature, June 2001, pp. 390–431.
10
Bruce C. Fallick and Charles A. Fleischman, “The Importance of
Employer-to-Employer Flows in the U.S. Labor Market,” Finance and
Economics Discussion Paper Series, no. 2001-18 (The Federal Reserve
Board, April 2001).
11
For more details, see Abraham, “Help-Wanted Advertising,”
1987, pp. 207–48; Bleakley and Fuhrer, “Shifts in the Beveridge Curve,”
1997, pp. 3–19; and Petrongolo and Pissarides, “Looking into the Black
Box,” 2001, pp. 390–431.
12
Abraham and Katz, “Cyclical Unemployment,” 1986, pp. 507–
22; and Blanchard and Diamond, “The Beveridge Curve,” 1989, pp. 1–
76.
13
For more details, see Abraham, “Help-Wanted Advertising,”
1987, pp. 207–48; Blanchard and Diamond “The Beveridge Curve,”
1989, pp. 1–76; and Bleakley and Fuhrer, “Shifts in the Beveridge
Curve,” 1997, pp. 3–19.
14
Abraham, “Help-Wanted Advertising,” 1987, pp. 207–48.
15
Abraham, “Help-Wanted Advertising,” 1987; and Blanchard and
Diamond, “The Cyclical Behavior of Gross Flows of Workers in the
U.S.,” Brookings Papers on Economic Activity, no. 2, 1990, pp. 85–155.
16
Fallick and Fleischman, “The Importance of Employer-to-Employer Flows,” 2001.
Monthly Labor Review
December 2001
37